Extraction you can check.
Paperloom turns PDFs, Office files, spreadsheets, and scans into structured, cited data. Every value links back to the exact spot on the original page — so you verify instead of trusting.
- Vendor…extracting
- Invoice #…extracting
- Issue date…extracting
- Line items…extracting
- Total…extracting
Hover a value to see exactly where it came from.
Audited 2026-08-04 against a private reviewed corpus, hash-stamped. Read the full audit
From dropped file to verified data.
No single model does everything well. Paperloom runs a pipeline that inspects, routes, extracts, and checks — and shows its work at every stage.
- 01 · Ingest
Upload anything
A single scan or a five-hundred-file batch. PDFs, Office files, spreadsheets, images — everything lands in one library, grouped and tracked.
- 02 · Classify
Classified in milliseconds
Every file is inspected — native text or scanned, simple or dense, one page or two hundred — before a single processing credit is spent.
- 03 · Route
Routed down one of four paths
Fast native extraction for clean PDFs. CPU layout analysis for structured files. GPU OCR for scans. Heavy vision models only when a document earns them.
- 04 · Extract
Extracted with coordinates
Fields, tables, and reading order come out with the exact region on the page each value was read from — provenance is part of the output, not an afterthought.
- 05 · Verify
Verified or flagged, never silent
Quality is scored on every run. Weak results escalate to a stronger path automatically; anything still uncertain becomes a reviewable exception instead of quietly passing.
The original and the answer, side by side.
Select an extracted field and Paperloom highlights the page and region it came from. Reading order, structured tables, and provenance stay visible in one working surface — no black-box output to take on faith.
4. Fees & Term
Payment terms
- Monthly fee
- $4,200.00
- Billing cycle
- Monthly · first business day
- Payment window
- Net 30
Select a clause. See its source.
Built for documents that matter.
Intelligent routing
Documents are classified by format and complexity, then processed on the cheapest path that can do the job — with quality-driven escalation when it can't.
Typed schemas & tables
Define versioned extraction schemas with typed fields and tables, draft one from a real document, diff versions, and roll back safely.
Exception triage
Missing fields, low confidence, and failed validation become reviewable exceptions with severity, evidence, and a clear resolution workflow.
Webhooks & API
Verified, review-required, and failed events delivered to your endpoints with signed payloads, per-attempt logs, and idempotent replay.
Measured accuracy
Extraction quality is audited against a reviewed corpus — exact values, exact table rows, citation coverage — and published, not promised.
Data controls
Versioned retention policies, on-demand deletion sweeps, and downloadable deletion certificates keep you in control of every stored page.
An API that shows its work.
Upload a document, get a signed webhook when extraction is verified — with a result you can fetch, and every value tied to its source. Signing secrets rotate, deliveries are logged per attempt, and events replay idempotently.
curl -X POST https://api.paperloom.dev/api/documents \
-H "Authorization: Bearer pk_live_…" \
-F "file=@northwind-invoice.pdf"{
"contract_version": "customer-webhook-event/v1",
"type": "extraction.verified",
"data": {
"document_id": "9c41…",
"schema_version": 3,
"extraction_status": "complete",
"reconciliation_status": "verified",
"result_url": "https://…/extractions/document/9c41…/latest"
}
}Stop re-keying. Start verifying.
Open source. Four processing paths. Every value cited back to its page.